Memory test
- Details
- Created: Monday, 06 May 2013 20:03
- Written by Kevin Figg
For one solid week, LV5 was left in a testing state running a Phantom network of 60 devices. Randomisation was set to a low-mid range value which means the network was not overly active. However, there were 343 rebursts over the course of the 7 days and each burst introduced 25 new devices instantly. Bursts are designed to emulate a collection of devices all powering on at the same time, or users logging in and out at the same time. Similar to how schools networks are.
The CPU usage was not measured for this early test, but instead, the memory usage was. At this point of the development of the server engine, it's important to know if there are any potential memory leaks.
The following graph shows the memory usage during this period. (Click for a larger version)
The data shows the servers memory usage is low in comparison to the Agent. Although the Agent itself is not final, it is actually the tool that creates the Phantom network, so I am not concerned by it's high memory usage. My focus is the server which is showing good memory management. It looks after the memory usage well, and over the course of the 7 days, it reduced fractionally. It also highlights that compared with the peaks of the Phantoms as shown by the Agent data, the server handles workloads very well. Given the server will be running 100% of the time, this is reassuring.
The testing done was by no means scientific. Normally a quick snapshot of the memory was done twice a day over the 7 days although occaisionally more was taken each day. The peaks show because at the time of the memory snapshot, a burst may have just happened or the station load was high. The agent will be tested at a later date when running as a service, and without running the Phantom network.